22results about How to "Reduce storage space requirements" patented technology

The invention belongs to the technical field of model training, and particularly relates to a large model increment training method and system based on dynamic sparsification. The method comprises the following steps: S1, evaluating the parameter importance of a large model by adopting a multi-dimensional evaluation method, and reflecting the contribution of parameters to the performance of the large model; s2, performing dynamic sparsification control on the large model structure by adopting a self-adaptive structure adjustment method; s3, performing incremental training management on the large model by constructing a capacity-adjustable empirical playback buffer pool and proposing a multi-objective loss function; and S4, carrying out model compression optimization on the large model by adopting a mixed precision quantification strategy. The method has the characteristics that the network structure can be dynamically adjusted by evaluating the importance of model parameters in real time, so that the efficient training is realized while the model performance is ensured.

The utility model belongs to the roll driving auxiliary structure technical field of roll grinder, especially relates to a numerical control roll grinder roll driving mechanism, include: positioning seat, connecting seat, gland and locking piece, be provided with positioning groove on the positioning seat, the positioning groove is the arc type structure that is suitable for the axle head outer profile of roll, the connecting seat sets up on the positioning seat, the connecting seat is coaxial distribution with the positioning seat, the positioning seat can be connected with the drive end on roll grinder head frame, the side of the gland close to the positioning seat is the arc type structure that is suitable for the axle head outer profile of roll, the gland can be along the radial of roll the axle head of roll is pressed in the positioning groove in the positioning seat, the locking piece can lock and unlock the gland with the positioning seat, compared with the prior art, the utility model solves the contact stress concentration, the back gap and inertia impact vibration and the higher problem such as cost of existing in prior art.

The utility model discloses a foldable circulating logistics box structure which comprises a bottom plate, vertical plates are symmetrically arranged on the upper portion and the lower portion of the two sides of the bottom plate, tongue plates are symmetrically arranged on the left side and the right side of each vertical plate, a supporting plate is arranged on the vertical plate located on the upper side, and an auxiliary supporting plate is arranged on the supporting plate. When the logistics box is used, all the parts are folded and spliced together, when the logistics box is not used, the logistics box is folded and unfolded to be in a flat plate shape, the requirement for storage space is greatly reduced, the cost and occupied space for transporting empty boxes are reduced, the logistics efficiency can be improved, and the box is usually made of recyclable materials such as polypropylene and is convenient to use. The plastic waste and environmental pollution can be reduced, through recycling, the requirement for new raw materials is reduced, the concept of sustainable development is met, the intelligent lock catch is used, the adhesive tape is prevented from being used, no discarded part is used, and the environment-friendly and recycling effects are achieved.

The utility model discloses a film-coated fluorine mets needled felt filter bag, which comprises a base cloth, a support mechanism and a protection mechanism, a fluorine mets needled felt layer is arranged outside the base cloth, a film-coated layer is covered outside the fluorine mets needled felt layer, a folding mechanism is arranged outside the base cloth and the film-coated layer, and the folding mechanism is arranged outside the base cloth and the film-coated layer. The folding mechanism comprises a first elastic supporting frame, and the surface of the first elastic supporting frame is fixedly connected with a first fixing plate. Through the arrangement of the supporting mechanism and the matching of the first elastic supporting frame, the first fixing plate, the first fixing hole, the second elastic supporting frame, the second fixing plate and the second fixing hole, the problem that a traditional filter bag is difficult to fold is solved, an operator can flexibly fold the filter bag into a proper size according to actual requirements, and the filter bag is convenient to use. The storage space requirement is lowered, the filter bags can be fixed through bolts after being folded, the filter bags are stacked in order, surface abrasion of the filter bags is reduced, the filter bags are convenient to take and use, and the time cost for overhauling production line equipment and replacing the filter bags is reduced.

The utility model provides a waste collecting device of a garment processing assembly line, and belongs to the field of garment processing waste collection, a controller is arranged on the front side wall of a collecting box, a discharging plate is arranged above the collecting box, a supporting frame is arranged at the top of a supporting table, a hydraulic cylinder is arranged at the top of the supporting frame, and the output end of the hydraulic cylinder is connected with an electromagnet. The compressing mechanism is arranged in the collecting box, and the compressing mechanism can extrude the waste in the collecting box into blocks; the vibrating mechanism is in linkage with the compression mechanism and is arranged on the side wall of the collection box, and the vibrating mechanism can drive the discharging plate to vibrate. According to the utility model, through the arrangement of the discharging plate and the compression mechanism, the discharging plate can guide waste materials on the conveyor belt into the collection box to be collected; and under the action of the compression mechanism, the waste materials entering the collecting box can be compressed, the size of the waste materials can be greatly reduced, the storage space requirement is remarkably reduced, and stacking and transportation are facilitated.

The utility model discloses an improved steel bar bending tool which comprises an operation table. The guide assembly is connected to one end of the upper portion of the operation table; the steel bar bending assembly is connected to the other end of the upper portion of the operation table. The transmission assembly is arranged on one side between the guide assembly and the steel bar bending assembly, one end of the transmission assembly is in transmission connection with the steel bar bending assembly, and the other end of the transmission assembly is connected to the upper portion of the operation table; and the operating handle is connected to the outer side of the other end of the transmission assembly. The steel bar bending machine is ingenious in design, safe and reliable, can accurately control the bending angle, ensures that steel bars meet design requirements, improves working efficiency, can provide stable and uniform power due to the arrangement of the transmission assembly, enables the bending process to be more stable, ensures that the bending angle and shape of each steel bar are consistent, improves consistency of bending quality, and reduces production cost. And meanwhile, the labor intensity is relieved, and fatigue and errors caused by manual bending are avoided.

The present invention relates to the technical field of electric meter data storage, and specifically discloses an electric energy meter data storage method and system based on user behavior analysis, comprising the following steps: step S1: setting a collection period, selecting and numbering a number of time nodes within the period; collecting users' electricity consumption time series data, and establishing an electricity consumption curve for each user; step S2: calculating the derivative value at each time node on the electricity consumption curve to obtain a derivative curve; performing cluster analysis on all derivative curves, selecting representative classes, and generating representative curves of the representative classes; step S3: for data to be stored, obtaining the derivative curve to be stored of the data to be stored; obtaining the similarity between the derivative curve to be stored and each representative curve, and judging whether the data to be stored is abnormal data; step S4: storing normal data in a regular storage space, and storing abnormal data in an abnormal data storage space.

This application provides a cleaning device, relating to the field of cleaning device technology. The cleaning device includes: a device body; a cloth assembly detachably connected to the device body, the cloth assembly including a traveling mechanism for driving the cloth assembly to be fixedly connected to or detached from the device body, and the traveling mechanism also for causing the position of the cloth assembly relative to the device body on the surface to be cleaned to change; and a connecting component including a first end and a second end, the first end connected to the device body and the second end connected to the cloth assembly, the distance between the first end and the second end being variable, the connecting component for restricting the movement of the cloth assembly within a preset area, the preset area being larger than the cleaning area covered when the cloth assembly is fixedly connected to the device body. The cleaning device provided by this application can solve the problem in related technologies where the cloth assembly loses its cleaning ability due to physical obstruction, resulting in poor cleaning effect.

The utility model relates to a multifunctional hanging seat for an overhead man-riding device, in particular to a hanging seat which is suitable for a coal mine overhead man-riding device and can rotate and be folded. According to the technical scheme, a rope clip is arranged on the top, a middle upper rod is arranged in the middle, a transverse handle is arranged on the upper rod, a lower rod facing front and back is arranged on the lower portion of the upper rod, a foot placing cross rod is arranged at the front end of the lower rod, a riding seat is arranged at the rear end of the lower rod, the bottom of the upper rod is connected with the top of a connecting shaft through a transverse hinged shaft, and a sleeve is arranged on the lower portion of the connecting shaft in a sleeved mode. The sleeve is welded to the side face of the lower rod, a supporting plate is arranged on the lower portion of the connecting shaft, a limiting groove is formed in the edge of the supporting plate, the rear end of the lower rod is connected with the mounting frame, and a turnover plate is hinged to the rear position, connected with the connecting shaft, of the lower rod and embedded into the limiting groove of the supporting plate to limit rotation between the lower rod and the connecting shaft. The lower rod and the upper rod of the device can be overturned, so that the passenger seat can be overturned upwards, the left and right side spaces are reduced, and storage management is facilitated.

This invention relates to the field of knowledge base technology, specifically to a knowledge base retrieval method supporting compression. It includes the following steps: S1, establishing a B+ tree as the index structure of the knowledge base, and compressing the B+ tree index structure using prefix compression technology; S2, compressing the data in the knowledge base using an intelligent differential compression algorithm; S3, adjusting the interaction between the index structure and compressed data using an intelligent index retrieval algorithm based on the B+ tree index structure; S4, monitoring performance indicators in real time and periodically checking the integrity of the data. This knowledge base retrieval method supporting compression optimizes data storage through an intelligent differential compression algorithm, significantly reducing storage space, and employs an intelligent index retrieval algorithm based on a B+ tree to achieve intelligent and fast data retrieval, optimize query paths, reduce decompression time, and improve retrieval efficiency.

The utility model discloses an arch-shaped combined keel flat-insertion partition wall system which comprises stand column keels, flat-insertion assemblies, panels, horizontal keels, inter-keel insertion fixing assemblies and panel supporting pieces, the backs of the panels are connected with the stand column keels in an inserted mode through the flat-insertion assemblies, the cross section of each stand column keel is of an arch-shaped opening structure, and each flat-insertion assembly comprises a mother piece and a son piece. The mother part is of a left-right symmetrical structure, a middle gap forms a child part insertion channel, the child part is a special-shaped L part and comprises child part parallel limbs and a child part insertion plate, a clamping protrusion is arranged at the head of the child part insertion plate, and the connecting concave position of the clamping protrusion and the child part insertion plate is limited through an end protrusion of a limiting hook. A closed stress ring is formed by flatly inserting the arch-shaped combined keel into the partition wall system, loads are effectively dispersed, and the overall stability is improved. The symmetrical angle plates of the female part are clamped into the inner clamping holes to form bidirectional constraint, the limiting hooks are matched with the clamping protrusions to achieve three-dimensional limiting, the loosening risk of traditional single-point insertion is avoided, and long-term stable connection of the panel is guaranteed.

The invention discloses a new energy project scheme multi-dimensional sensitivity problem rapid analysis method, and relates to the technical field of data analysis, and the method comprises the steps: inputting the basic data of a new energy project; mapping the multi-dimensional sensitive parameters of the new energy project to quantum bits; and performing parallel calculation on the quantum bits through quantum state superposition to obtain an analysis result under the multi-dimensional sensitive parameter combination. According to the method, the sensitive parameters are mapped into the quantum bits, and the superposition characteristic of the quantum state is utilized, so that parallel calculation of all parameter combinations can be realized, the calculation efficiency is greatly improved, and the resource consumption is reduced. By using the parallelism of quantum calculation, a large number of variable combinations can be processed simultaneously in one evolution process, and the exponential growth problem of a traditional permutation and combination method is avoided. The calculation cost is reduced, the operability and the application value of the system are improved, and the application scene of multi-dimensional sensitivity analysis is expanded.

The application provides an efficient matrix storage method and system based on wind power prediction data, and relates to the technical field of wind power data storage. The method comprises the following steps: constructing multi-dimensional time series data according to wind power prediction data of each wind turbine in a wind farm; constructing a table structure based on an array field for the multi-dimensional time series data, each row corresponding to prediction data of a prediction model of the wind turbine at a prediction release time, and each feature storing the numerical value of the corresponding feature at multiple prediction target times in the form of an array field; performing hierarchical storage management on the wind power prediction data according to the table structure, storing wind power prediction data within a first time range as hot data in a cache table in the form of an original array; and storing the remaining wind power prediction data as cold data, obtaining multiple time blocks, performing singular value decomposition on the original matrix corresponding to each time block, and compressing and storing the cold data into a cold data table. The application helps to improve data calling efficiency and reduce overall storage space occupation.

The present invention provides a vehicle-mounted launch device and method for a drone, belonging to the technical field of drone launch. The vehicle-mounted launch device for a drone includes: a lower frame; an upper frame, the upper frame is connected to the lower frame through a connecting rod; a support leg, the support leg is fixedly connected to the lower end of the lower frame; two limit plates, both of which are fixedly connected to the upper end of the upper frame; a limit rod, the limit rod is arranged in the two limit plates; the angle of the upper frame can be adjusted according to actual needs through an angle adjustment mechanism, so as to provide an optimal launch posture for the drone; the drone can be quickly deployed under different terrain conditions, and the launch angle can be flexibly adjusted according to different mission requirements, thereby increasing the adaptability and scope of use of the device. In addition, the vehicle-mounted design allows the entire launch device to be easily transported to any location, enhancing the maneuverability and response speed of the operation.

The application provides a temperature field prediction method and medium, comprising: obtaining a temperature field data set of a rocket engine; inputting the temperature field data set into a pre-trained temperature field prediction model to obtain temperature field prediction time series data in a prediction time period; wherein the temperature field prediction model comprises sequentially connected time series compression module, time series prediction module, time series decompression module and time series output module; the time series compression module is configured to compress the temperature field time series data to obtain temperature field compressed time series data; the time series prediction module is configured to perform prediction according to working condition data and the temperature field compressed time series data to obtain compressed prediction data in each prediction time of at least one prediction time; and the time series decompression module is configured to decompress the compressed prediction data in each prediction time to obtain temperature field prediction data in each prediction time. In this way, rapid and accurate prediction of the temperature field of the rocket combustion chamber can be realized.

The invention discloses a data averaging algorithm for pretightening force electromagnetic ultrasonic measurement, which comprises the following steps of: (1) converting Nth frame echo data received by an electromagnetic ultrasonic probe (i) into digital signal data D through an amplifier and an analog-to-digital converter (d), representing the digital signal data D as # imgabs0 #, representing the number of data points received by each frame of probe, and representing the number of data points received by each frame of probe; the # imgabs1 # is decomposed into effective echo data # imgabs2 # and random noise # imgabs3 #, and digital signal data # imgabs4 # are obtained; (2) dividing the historical accumulated data A stored in the storage space (a) into two paths of data; according to the data averaging algorithm, only the current frame echo data and the historical accumulated data need to be processed each time, the historical accumulated data received by the electromagnetic ultrasonic probe only needs to be stored once in the storage space, and the average data of electromagnetic ultrasonic measurement can be output after new frame echo data are received each time, so that the calculation amount is greatly reduced, and the calculation efficiency is improved. The data processing speed is improved, the repetition frequency of data output is reduced, the signal-to-noise ratio is higher, and pretightening force measurement is facilitated.

The invention discloses a program upgrading method and device, a vehicle and a storage medium. The program upgrading method comprises the steps that a compressed upgrading file sent by remote equipment is received; writing the compressed upgrading file into a backup area corresponding to the controller; decompressing the compressed upgrading file in the backup area to an operation area corresponding to the controller; and performing program upgrading on the controller according to a decompressed file obtained by decompressing the compressed upgrading file in the running area. Thus, the upgrading file is compressed in advance, the controller receives the compressed upgrading file obtained through compression and directly writes the compressed upgrading file into the backup area corresponding to the controller, the requirement for the storage space of the backup area is reduced, meanwhile, the time needed for file transmission and writing is shortened, and therefore the program upgrading speed is greatly increased.

This invention relates to a time-refreshing method for hardening missile-borne computers against single-event upsets (SWEs), belonging to the field of missile-borne computers. In this invention, a DSP processor (1) and a non-volatile memory (2) are connected via an EMIF bus (3). Software hardened using the SEU method runs in the DSP processor. The non-volatile memory is responsible for backing up the protected data, and the EMIF bus is responsible for supporting data interaction between the DSP processor and the non-volatile memory. This invention does not increase the internal storage space requirements of the DSP; SEU hardening is performed at a certain time period, allowing for time-sharing of DSP computing resources, thus minimizing the impact on normal program execution; the length of data blocks and the cycle of SEU hardening can be flexibly adjusted, achieving a balance between computing resource consumption and SEU hardening effectiveness.

The application provides a display device and a driving method thereof. The driving method comprises the following steps: measuring a high level duration of a frame start signal of the display device; calculating a current driving frequency of the display device based on the high level duration; updating a color coordinate accuracy lookup table of a predetermined target gray scale when the current driving frequency is lower than a preset threshold; and adjusting the brightness of the display device according to the updated color coordinate accuracy lookup table. The application can greatly reduce the storage space requirement of the compensation data in the timing controller by selectively updating the color coordinate accuracy lookup table, and ensure the smoothness of the brightness adjustment by updating the adjacent gray scale data in a gradient manner. In addition, the application does not require additional hardware support, thereby effectively reducing the cost. Experimental results show that the application can significantly improve the variable refresh rate performance, effectively reduce the brightness flicker, and improve the display quality.

The utility model discloses a roof covering all-in-one machine which comprises a base, a rotary support is connected to the base in a rotating mode, a rotary disc is installed at the upper end of the rotary support, a plurality of rollers are arranged on the rotary disc at equal intervals, the rollers are connected through a conveying motor and a chain, a rotary motor is installed in the rotary disc, and the rotary motor is connected with the rotary support in a rotating mode. An output shaft of the rotary motor is connected with the rotary support through a gear, a machine frame is installed on one side of the upper end of the base, a supporting frame is installed on the machine frame, and a roof covering lifting mechanism is installed on the supporting frame. Multiple devices do not need to be used in cooperation, the occupied area of the devices is greatly reduced, the requirement of enterprises for storage space is lowered, the manual operation time and workload are reduced, a large number of goods are packaged within a short time, the production efficiency of the enterprises is improved, and the logistics turnover speed is increased. Packaging parameters can be flexibly adjusted according to the sizes, shapes and packaging requirements of different goods, automatic production is achieved, and packaging adaptability is improved.

The utility model relates to the technical field of electric tools, in particular to a mold quick-changing structure of an electric tool, which comprises an upper mold assembly, a lower mold assembly and a lower mold assembly, the lower die assembly is composed of a lower main die and a replaceable lower auxiliary die in a cooperative mode. The two parts are combined for use, so that accurate matching is provided for different workpiece specifications, and the flexibility in module operation is ensured; the die holder assembly is composed of an upper die holder and a lower die holder and used for wrapping the upper die assembly and the lower die assembly, so that the stability between the upper die assembly and the lower die assembly is improved; the locking assembly comprises a first rack, a second rack, a gear and a reset spring. Compared with the prior art, the mold system has the advantages that by adopting the design of the replaceable upper auxiliary mold and the replaceable lower auxiliary mold, the same set of mold system can be compatible with various battery pack specifications, the number of molds needing to be independently designed and manufactured is greatly reduced, the production cost is remarkably reduced, the mold preparation period is shortened, and the production efficiency is improved. And the production efficiency and the market response speed are improved.

The application discloses a cross-document type retrieval method and system based on differential nested coding. The method includes six steps: document preprocessing, document type identification, differential coding, index construction and optimization, online retrieval and result fusion. First, the document is standardized and preprocessed to extract structured information. Second, the BERT classifier is used to analyze the document features and identify the document type. Then, according to the document type, the corresponding encoder is selected to generate the original coding vector. Next, the differential nested coding index is constructed and optimized. In the online retrieval stage, the user query is preprocessed, the relevant document type is predicted, and the approximate nearest neighbor retrieval is performed. Finally, the retrieval results are sorted by a stepwise weight fusion strategy, and the top-100 retrieval results are returned. Through differential coding and index optimization, the application realizes efficient retrieval of different types of documents, and has significant practical value and application prospect.